Research Scientist — Scaled RL Infrastructure

Openkrill; United States; Remote; USD 230,000 - 300,000; Full time

Job summary

Luma is seeking a senior RL engineer to build and scale post-training systems that couple policy optimization with large fleets of inference workers and agentic environments. You will streamline the full RL loop across thousands of GPUs, ensuring speed, stability, and correctness in training, rollout, and reward computation.

The role entails designing environments, reward infrastructures, and evaluation tooling, plus advancing training efficiency and deploying new ideas into production.

Qualifications

• Experience with post-training LLMs using RL (PPO/GRPO-family, RLHF, RLVR).
• Strong background in distributed training and parallelism for foundation models.
• Experience building RL environments, reward functions, verifiers, or evaluation harnesses.

Responsibilities

• Design, build, and scale distributed RL post-training systems across thousands of GPUs.
• Build high-throughput rollout generation and integrate inference engines and weight synchronization.
• Design RL environments for agentic, multi-step tasks with sandboxed execution and tool use.
• Build reward infrastructure: verifiable rewards, reward-model serving, LLM-as-judge pipelines.
• Develop evaluation, monitoring, and debugging tooling for large RL runs.
• Advance training efficiency and stability, turning post-training ideas into production.

Skills

RL post-training

Distributed PyTorch

RL environments

RL post-training frameworks

NCCL/MPI

Tools

vLLM

SGLang

veRL

OpenRLHF

TRL

Ray

Job description

Luma is seeking a senior RL engineer to build and scale post-training systems that couple policy optimization with large fleets of inference workers and agentic environments. You will streamline the full RL loop across thousands of GPUs, ensuring speed, stability, and correctness in training, rollout, and reward computation.

The role entails designing environments, reward infrastructures, and evaluation tooling, plus advancing training efficiency and deploying new ideas into production.